That sort of countermeasure system could be done without AI as well. The problem is that it's illegal. No Castle Doctrine in cybersecurity afaik.

Interesting variation on that could be AI that builds out some sort of on the fly honeypot after identifying the attacker. Basically creating the "attack" within their own premises.